From the moment of its inception, artificial intelligence (AI) has been woefully misunderstood. On the one hand, you have fears surrounding the looming, ever-present singularity, the idea pushed by the media that machines will one day rise up and overthrow us. On the other, you have journalists and tech bloggers who look at AI and promise the sun, moon, and stars.
Then you look at what AI can actually do.
Not to say that the technology doesn’t have promise. Cognitive computing has the potential to completely change how we live and work, and in many ways is already doing so. In order to realize that potential from a cybersecurity perspective, you need to look past the misconceptions around the technology.
One of the most common we’ve encountered involves the conflation of AI with machine learning (ML).
What Is Artificial Intelligence?
AI is the capacity of a computer or endpoint to mimic human cognitive function. This allows a system to perform a range of tasks that would ordinarily require human intervention. Said tasks can be either simple or complex in nature.
Per Government Technology, there are four types of AI:
- Reactive. Capable of drawing on a vast pool of data to evaluate a scenario, but can neither learn nor draw on past experiences. The supercomputer Deep Blue is an example of reactive AI.
- Limited Memory. Capable of drawing on stored data to inform future decisions through comprehensive training or real-world experience. Many chatbots leverage limited memory AI.
- Theory of Mind. A theory of mind AI is capable of recognizing other entities around it, and incorporating their possible experiences, thoughts, emotions, and behavior into its decision-making. This is the peak of current AI technology, and we have only achieved rudimentary progress in this regard.
- Self-Awareness. AI that is capable of forming conscious thoughts and a concept of self. We are currently nowhere close to this.
What Is Machine Learning?
ML refers to the processes and techniques by which limited memory (and potentially theory of mind) AI is trained. It’s the foundation that makes more advanced AI systems possible and primarily digests structured data. Deep learning, as defined by IBM, is a more advanced form of ML that uses neural networks to analyze unstructured data.
How Are AI and ML Applied in Cybersecurity?
ML plays a crucial role in training AI-driven cybersecurity solutions to recognize common threat patterns. Most malicious software tends to behave in a very similar fashion. ML also allows smart security tools to establish a baseline for what constitutes ‘normal’ down to a granular level.
For instance, if a remote employee always logs in from Seattle but then attempts to connect to corporate systems from Russia one day, ML would allow a security solution to recognize this as a potential attack flag.
Smarter Security Starts Here
AI and ML are ultimately two sides of the same coin. ML is required for more advanced forms of AI to exist. And without an AI platform to feed training data, ML provides little in the way of value.